Date: Thu, 28 Feb 2013 09:36:20 +0400 From: rihad <rihad@mail.ru> To: John Baldwin <jhb@freebsd.org> Cc: freebsd-stable@freebsd.org, vince@unsane.co.uk Subject: Re: mfi timeouts Message-ID: <512EECD4.5080600@mail.ru> In-Reply-To: <201302271159.13424.jhb@freebsd.org> References: <512CFF90.8080806@mail.ru> <201302261548.56253.jhb@freebsd.org> <512DA073.9090908@mail.ru> <201302271159.13424.jhb@freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 02/27/2013 08:59 PM, John Baldwin wrote: > On Wednesday, February 27, 2013 12:58:11 am rihad wrote: >> Now about this part taken from here >> http://lists.freebsd.org/pipermail/freebsd-scsi/2011-March/004839.html >> > By issuing a dummy read operation (thus forcing a flush of data >> buffers), this issue is largely averted. >> >> Does this mean that battery-backed cache (BBU) is effectively rendered >> useless, as all write operations are forced on to the disk platters on >> every interrupt? > No, this is a very different level. This is forcing pending PCI DMA > transactions on the PCI bus to flush by doing a read, not forcing I/O > buffers to be flushed to disk. > Thanks for clarifying. After applying the "dummy read" patch mfi timeouts don't appear in dmesg output any more, but i/o stalls still occurred 2-3 times during periods of high activity, for no more than 10-20 seconds. I guess the only way to fix that is to choose another hardware RAID implementation, or try Steven Hartland's patch? Does 8.3 or 9.1 include more fixes in this area, is upgrading recommended?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?512EECD4.5080600>